 So in its current state, you have everything you need to be able to train and test and evaluate a convolutional neural network on your whatever computer you're holding whether it's a desktop or a laptop or something else This is set up cottonwood is designed to run on your local machine All you need to have is Python running This is not going to be as fast as something that's running on GPUs. It's been optimized It probably won't be as fast even as PyTorch or TensorFlow running on your laptop. It is not optimized for speed. It's not a race car It's not designed to go as fast as possible It's more like a go-kart that you can see all the way through to the engine. There's no body body panels you'll see when we go to look at the code that Everything is designed to be as transparent as possible as easy to see how it works as we can make it this comes often at the expense of Performance so there are ways that we could speed things up or optimize things that we don't do because it makes the code harder to Understand or because it requires incorporating other large libraries or programming Practices that make the code much larger. The goal here is to have a teaching tool that lets you see all the way through to the middle So that when you go to implement this on a larger established framework You have a gut feel for what's going on and when you're in there tweaking parameters and changing things You can make sure that you aren't doing it just on a whim That you have a sense of why you need to change things and how they should be changed to get the results that you want The other thing that cottonwood gets you is because it is simple. It's like the Lego block architecture level of complexity if you have a new idea For something that you want to stick in the middle you want to try out a new concept or a new implementation of something You can code it up and integrate it relatively Easily it takes a lot less work than it might do to write some custom code from scratch and put it into a pytorch network and This is also meant to Shorten the self teaching time the hey, I bet I could do this differently Let me change this thing and see how it works. Oh, that did not work the way I thought it would or oh my god, that works amazing Let me take a step back now think about how I'd want to scale that up Implement it in pytorch analyze it, but it lets you close that first iteration loop very quickly and Freeze you up to explore ideas rapidly So it's supposed to be optimized for teaching teaching others or teaching yourself